Undercut

Undercut forms when the arc melts more parent material at the edge of the joint than the filler metal arriving behind it can fill back in. The result is a groove running along the weld toe — occasionally along the root on the reverse side of a single-sided weld — that is real base-metal loss, not a coating or a shadow, and it does not fill itself in on cooling.

The mechanism is almost always an energy or speed mismatch: current or arc voltage above the qualified range washes material out of the toe faster than the pool can wet and back-fill it, and high travel speed compounds it by leaving the pool less time to flow into the groove before it solidifies. Wide weaving and steep torch angles can produce the same effect locally even when average parameters are within range, which is why undercut can appear intermittently along an otherwise well-controlled weld.

Its consequence is disproportionate to how little metal is actually missing. A shallow groove at the toe is a sharp geometric notch sitting exactly where service stress in a fillet or butt weld is already concentrated by the change in section, so undercut is treated first as a fatigue-crack initiation site and only secondarily as a cosmetic or dimensional issue. ISO 6520-1 classifies continuous and intermittent forms of the imperfection separately, because a continuous groove running the length of a weld behaves differently under load than isolated short ones.

Because the groove is on the surface, direct visual inspection remains the leading detection route, unlike planar imperfections such as lack of fusion that hide inside the joint. Repair by grinding to a smooth, blended profile is common, but grinding removes material from the same weld toe region and, if it is not controlled, can locally reduce the effective throat thickness of a fillet weld below what the repair was supposed to restore.